交通监控视频的快速图像稳定方法

Fast Image Stabilization for Traffic Surveillance Video

安装在户外的摄像机容易受到各种干扰,因此图像稳定是交通视频监控过程中必不可少的一步。本文采用背景图像中含有较大曲率边缘的区域做运动估计,并判别出全局运动矢量。再根据一个简单的3参数摄像机运动几何模型,估算出全局运动参数,进行全局运动补偿达到快速图像稳定。该方法结合特征匹配方法与运动判别方法的优点,模型参数少,求解过程快,能很好满足监控应用中图像稳定的实际需求,也可以用来对驾驶疲劳警示和车辆碰撞预防等场合获取的视频图像实现稳定化处理。

The camera mounted outdoors is very easy to be disturbed;hence image stabilization is an essential step in traffic surveillance. This paper utilizes the area contains higher curvature edge in the background to estimate the local motion and extract the global motion vectors by motion decision, then proposes a novel three-parameter geometrical model to estimate the global motion parameters. Fast image stabilization will be accomplished by global motion compensation utilized the novel global motion model, This method combines the merit of feature matching and motion decision, and needs less computational cost because the model only has 3 parameters, so it can work well in actual application. It is also helpful for the image stabilization of driving fatigue detection and vehicle crash prevention.